GPT-4o

Can we talk about GPT-4o? I tried something I read in an article because I didn’t believe it was true. It actually worked flawlessly and I’m gobsmacked.

I uploaded an image of the Atari game breakout and asked GPT-4o to code me a playable version of the game based on the screenshot. It 0-shot an application that worked right out the gate in under 2 min.

Think about the implications of that. This is obviously a very simple example, but imagine the ability to reverse engineer almost any capability your competitor has, that you wish had, just by grabbing screenshots from their platform…

In 5 years software companies will be largely defunct unless they have a moat in the form of

a) high infrastructure cost as a barrier to entry (Gen AI companies themselves come to mind)
b) owned and protected data (note this means it can’t be accessible to AI models especially)
c) network effect (social platforms for example would be hard to replicate the user base)

I would go further to say that applications will become more and more personalized, similar to what’s happened with the democratization of content creation. I will have my suite of personal apps for ME and MY way of working, and you will have yours. My personal competitive differentiator will be how effective I’ve been at creating this personalized suite to enhance my productivity relative to my peers with their personalized suite of tools.

It’s going to be whole new ball game, and I’m here for it.